Trusted Uncertainty in Large Language Models: A Unified Framework for Confidence Calibration and Risk-Controlled Refusal
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866914222746632192 |
|---|---|
| author | Oehri, Markus Conti, Giulia Pather, Kaviraj Rossi, Alexandre Serra, Laia Parody, Adrian Johannesen, Rogvi Petersen, Aviaja Krasniqi, Arben |
| author_facet | Oehri, Markus Conti, Giulia Pather, Kaviraj Rossi, Alexandre Serra, Laia Parody, Adrian Johannesen, Rogvi Petersen, Aviaja Krasniqi, Arben |
| contents | Deployed language models must decide not only what to answer but also when not to answer. We present UniCR, a unified framework that turns heterogeneous uncertainty evidence including sequence likelihoods, self-consistency dispersion, retrieval compatibility, and tool or verifier feedback into a calibrated probability of correctness and then enforces a user-specified error budget via principled refusal. UniCR learns a lightweight calibration head with temperature scaling and proper scoring, supports API-only models through black-box features, and offers distribution-free guarantees using conformal risk control. For long-form generation, we align confidence with semantic fidelity by supervising on atomic factuality scores derived from retrieved evidence, reducing confident hallucinations while preserving coverage. Experiments on short-form QA, code generation with execution tests, and retrieval-augmented long-form QA show consistent improvements in calibration metrics, lower area under the risk-coverage curve, and higher coverage at fixed risk compared to entropy or logit thresholds, post-hoc calibrators, and end-to-end selective baselines. Analyses reveal that evidence contradiction, semantic dispersion, and tool inconsistency are the dominant drivers of abstention, yielding informative user-facing refusal messages. The result is a portable recipe of evidence fusion to calibrated probability to risk-controlled decision that improves trustworthiness without fine-tuning the base model and remains valid under distribution shift. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_01455 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Trusted Uncertainty in Large Language Models: A Unified Framework for Confidence Calibration and Risk-Controlled Refusal Oehri, Markus Conti, Giulia Pather, Kaviraj Rossi, Alexandre Serra, Laia Parody, Adrian Johannesen, Rogvi Petersen, Aviaja Krasniqi, Arben Computation and Language 68T50 I.2.7 Deployed language models must decide not only what to answer but also when not to answer. We present UniCR, a unified framework that turns heterogeneous uncertainty evidence including sequence likelihoods, self-consistency dispersion, retrieval compatibility, and tool or verifier feedback into a calibrated probability of correctness and then enforces a user-specified error budget via principled refusal. UniCR learns a lightweight calibration head with temperature scaling and proper scoring, supports API-only models through black-box features, and offers distribution-free guarantees using conformal risk control. For long-form generation, we align confidence with semantic fidelity by supervising on atomic factuality scores derived from retrieved evidence, reducing confident hallucinations while preserving coverage. Experiments on short-form QA, code generation with execution tests, and retrieval-augmented long-form QA show consistent improvements in calibration metrics, lower area under the risk-coverage curve, and higher coverage at fixed risk compared to entropy or logit thresholds, post-hoc calibrators, and end-to-end selective baselines. Analyses reveal that evidence contradiction, semantic dispersion, and tool inconsistency are the dominant drivers of abstention, yielding informative user-facing refusal messages. The result is a portable recipe of evidence fusion to calibrated probability to risk-controlled decision that improves trustworthiness without fine-tuning the base model and remains valid under distribution shift. |
| title | Trusted Uncertainty in Large Language Models: A Unified Framework for Confidence Calibration and Risk-Controlled Refusal |
| topic | Computation and Language 68T50 I.2.7 |
| url | https://arxiv.org/abs/2509.01455 |